Archived

This topic is now archived and is closed to further replies.

Fill rate performance

This topic is 5273 days old which is more than the 365 day threshold we allow for new replies. Please post a new topic.

If you intended to correct an error in the post then please contact us.

Recommended Posts

I''m writing an engine and I''m wondering if I need to make some optimizations. If a polygon is completely outside of the clipping frustrum, it isn''t going to take up much time in terms of the fill rate, right? Do modern video cards check if a polygon is outside the viewing frustrum before starting to draw? So if I send 5000 polygons to the hardware, with only 500 onscreen, how big of a hit do I take from the 4500 other polygons? I understand that this is likely card specific...

Share this post


Link to post
Share on other sites
theoretically it should take almost no time for the graphics card to drop not visible polys but you need to pass all these vertices to the card, transform them,...

guess why modern graphics engines take that long to develop...
(why do you think doom3 looks muuuchh better than most of the software that is currently out there (realtime that is) if it is not for discarding every possible calculation possible...)

mfg Phreak
--
"Input... need input!" - Johnny Five, Short Circuit.

Share this post


Link to post
Share on other sites
i think this depends on whether u can tell easily whether the polys are in the fov before u send them off for rendering... for example, if u have a city that''s divided into blocks, and u know that from block #1 there is no way u can see block #17, then there''s no reason to send block 17 polys to the video card. however, if u''re unsure whether a poly is visible, or just outside ur fov, then i suggest you send the polys to the video card and let that decide whether that poly needs to be rendered (ie. the card should see not, and thus not apply, say, bump-mapping on it).

this just what i think, so i could be wrong, since i''ve never done anything major in 3d (yet).

perhaps this''ll serve as a bump so that someone else more educated on the topic can help you.

---
shurcool
wwdev

Share this post


Link to post
Share on other sites
it would depend on the order.. if its outside the frustum you "only" waste the time for sending and transforming the geometry. if its occluded like in the city example it is sorted out if the parts closer to the cam have already been drawn (but on a per pixel base id expect). if you render them first then you have to do all the work.

or short: if you need occlusion culling depends on your scene, but dont even think about doing complex scenes without frustum culling ,-) (though im talking more about ten thousands of polys).

i better another thing: dont think about doing culling per polygon. do it per object and try to get hundreds or thousands at once by testing a bounding volume. use the knowledge about the scene/world that your card doesnt have ,-)

[edited by - Trienco on July 7, 2003 11:59:33 AM]

Share this post


Link to post
Share on other sites